introduction to swarm intelligence

18
SWARM INTELLIGENCE SWARM??????

Upload: aishwarya-rajkumar

Post on 21-Apr-2017

101 views

Category:

Education


1 download

TRANSCRIPT

SWARM INTELLIGENCE

SWARM??????

What for????• Capable for producing low cost• Fast• Robsut solution to several complex problem• Simple rule for each individual• Decentralized• Emerget

Swarm intelligence

Proposed by: James kennedy and Russell eberhart in 1995

Inspired from the natural social behavior and dynamic movements with communication of birds , insects and fish.

RULESRule 1: Separation Rule 2: Alignment

Rule 3: Cohesion

Basic particle swarm optimization

In computer science, particle swarm optimization (PSO) is a computational method that optimizes a problem by iteratively trying to improve a candidate solution with regard to a given measure of quality.

• Search area(possible solution)• Moving towards the promising area• Global best(each particles)• Each particles modifies its position according

to, current position current velocity distance between current position and pbest distance between current position and gbest

Algorithm parametersA:population of agentsPi:position of agents ai in the solution spacef :objective functionVi:velocity of agents ai

V(ai):neighbourhood of agents ai(fixed)

Algorithm[X*]=PSO()P=Particle_Intialization();for i=1 to it_maxFor each particle p in P dofp=f(p);If fp is better than f(p Best)pBest=p;endendgBest=best p in P;For each particle p in P doV=V+C1*rand*(pBest=p)+C2*rand*(gBest-p);P=p+V;endend

Particle Update Rulep=p+V with V=V+C1*rand*(pBest=p)+C2*rand*(gBest-

p);Where,p=particles positionV=path directionC1=weight of local information(personal best value)C2=weight of global information(neighbor best value)pBest=Best position of the particlegBest=Best position of the neighbor rand=Random variable

Update particle velocity

Vi ^t+1=Vi^t+C1V1^t(pbi^t=pi^t)+C2U2^t(gbi^t-pi^t)Diversification Intensification

Inertia Personal influence

Social influence

Makes particles moves in the same direction and

same velocity

Improves the individual makes the particle return from the current position

Makes the particle to follow the best solution

Advantages and Disadvantages

Few algorithm parametersEfficient global search algorithmSimple implementationEasily parallelized for concurrent programmingSlow convergence

Thank you